Descriptions:
Greg Isenberg interviews Alli K. Miller — former IBM and AWS executive who managed multi-billion-dollar P&Ls in AI — for a practical conversation on what it actually takes to build and operate an AI agent workforce at scale.
Miller’s central argument is that “managing agents” is the wrong mental model. The real role, she says, is closer to an SVP setting infrastructure and waiting for escalations than a direct manager assigning tasks. She runs a 34-agent workforce with an AI chief of staff named Simon, and describes her highest-performing prompt as just three words — framing it as setting ambitious ceilings rather than micromanaging execution. The progression she recommends for getting started: first work with one reactive agent, then a proactive one, then two agents routing to each other, then a full workforce with a mission control view showing how context passes between agents.
Practical design principles discussed include assigning traditional job titles to agents (CMO, CPO) to frame responsibilities clearly, using lighter models like Claude Haiku and Sonnet for sub-agents rather than defaulting everything to Opus, and deploying AI watchdogs in Slack to catch duplicative work and surface blockers. Miller also shares a one-prompt bootstrapping approach for founders — describing company context, goals, and constraints in a single prompt to spin up an initial workforce quickly before refining through iteration. The episode is a useful reference for anyone moving from individual AI tool use toward coordinated multi-agent deployments.
📺 Source: Greg Isenberg · Published August 12, 2026
🏷️ Format: Podcast
